Release 3.5.1
KubeMQ 3.5.1 makes Kafka share groups generally available with Kafka 4.3 parity and adds kmq update and kmq version --check.
Released September 27, 2026. Kafka share groups (KIP-932) are generally available, and kmq can update itself. The operator and the Helm chart are unchanged from 3.5.0.
Artifacts
| Artifact | Version |
|---|---|
| kmq | v3.5.1 |
| Server image | europe-docker.pkg.dev/kubemq/images/kubemq-next:v3.5.1 |
| Server image (FIPS) | europe-docker.pkg.dev/kubemq/images/kubemq-next-fips:v3.5.1 |
| Operator image | europe-docker.pkg.dev/kubemq/images/kubemq-operator-next:v3.4.0 (unchanged) |
Helm chart kubemq-next | 3.4.0 (unchanged; it installs server v3.5.0) |
Before you upgrade
-
Kafka share groups: upgrade every server before share traffic resumes. This release adds new cluster commands for share groups, controlled by
CONNECTORS_KAFKA_SHARE_FEATURE_VERSION(default2). A server still on an older release stops when it meets one of them, and keeps stopping until it is upgraded; on a 3-server cluster this can cost the cluster its majority. Before a rolling upgrade, do one of these:- Stop share consumers and per-group configuration changes until every server runs 3.5.1, then upgrade normally.
- Set
CONNECTORS_KAFKA_SHARE_FEATURE_VERSION=1on each server as you upgrade it, and raise it to2, with a restart, once every server is upgraded. At1, the new share-group features are unavailable: share groups cannot be deleted, per-group configuration writes andread_committedshare groups are refused, and lock renewal is off.
Downgrading is supported only if the setting stayed at
1the whole time. Do not setshare.isolation.level=read_committeduntil every server runs 3.5.1. -
New share groups start at the latest offset, as in Apache Kafka. Existing groups keep their position. To start new groups at the earliest offset, as before, set
CONNECTORS_KAFKA_SHARE_AUTO_OFFSET_RESET=earliest. -
Share groups count against
CONNECTORS_KAFKA_MAX_GROUPS, the limit that already applies to classic consumer groups. A share group counts untilDeleteGroupsremoves it. -
Rolling restarts count a delivery attempt. Every share record in flight when a partition's leader changes counts one delivery attempt, including on each server restart of a rolling upgrade. Keep the per-group delivery limit at the default of 5 or higher.
Changes
- Kafka share groups are generally available, certified with the Java
KafkaShareConsumer4.3 and the Kafka 4.3 console tools on a single server and on 3-server clusters. See Share groups. - Share-group administration:
ListGroupswith the type filter,DeleteGroupsonce no consumer is attached, per-group settings throughkafka-configs.sh --entity-type groups, andread_committedshare groups. - Share-group monitoring: the
kubemq_kafka_share_group_lag,kubemq_kafka_share_group_in_flightandkubemq_kafka_share_group_redeliveredmetrics,GET /api/kafka/share-groups, share groups on the dashboard's consumer-groups tab, andkmq kafka share-groups. - Kafka fixes: a node that became the Kafka coordinator right after it became ready now loads the committed offsets, and a clustered topic-configuration reload that fails no longer serves empty settings or loses an acknowledged change.
- kmq updates itself.
kmq updateinstalls the latest kmq release, andkmq version --checkreports whether a newer one exists. On Windows the update is scheduled rather than applied at once.
Known issues
- Images are published for
linux/amd64only. Release 3.5.2 addslinux/arm64. - No default connector image is published. Set
spec.imageon everyKubemqConnector.
Was this page helpful?
Release 3.5.2
KubeMQ 3.5.2 publishes server images for arm64 as well as amd64, lowers the management password minimum to 6 characters, and clarifies kmq sign-in.
Release 3.5.0
KubeMQ 3.5.0 lets kmq install and license KubeMQ, supports single-server Podman and HTTPS management, and splits the Helm operator and cluster releases.